Oxygen vacancy (V_O)-associated defects are often considered as active sites that are composed of unsaturated coordination structures in oxide catalysts in catalytic reactions. The concentration of V_O in metal oxides is usually determined by the oxygen deficiency, which is normally dynamic and controlled by the properties of the electron orbitals of cations in the oxides [1–5]. Therefore, the V_O needs to be uniformly distributed and effectively stabilized in the crystal lattice. Creating V_O in the catalysts would highly facilitate the catalytic activity, while stabilizing the V_O would improve the durability of catalysts [6,7].

β-Ga₂O₃ is a highly efficient catalyst that is usually used for light alkane dehydrogenation reaction [8,9]. β-Ga₂O₃ is a monoclinic n-type semiconductor with the space group of C₂/m and the lattice parameters a, b, c, and β of 12.23 Å, 3.04 Å, 5.80 Å, and 103.7°, respectively [10,11]. The unit cell consists of 2 Ga ions and 3 types of O ions that are located in the tetrahedral and octahedral coordination, respectively. Two O ions have a triple coordination, and the third O has a tetrahedral coordination. Due to the unique feature of electron orbital of Ga, the β-Ga₂O₃ is chemically and thermally stable especially at reduced temperatures. Reduction treatment produces V_O in β-Ga₂O₃ at a certain temperature, which creates the low-coordinated Ga active sites in the electron-deficient surface. It has been confirmed that these active sites markedly facilitate the enhancement of catalytic activity of nonoxidative C₂H₆ dehydrogenation, especially for the C–H activation, but avoid the C–C bond breakage at reduced temperatures [12–14]. C₂H₄ is the most widely used block material in the manufacture of polymers and other chemicals [15,16]. Although the β-Ga₂O₃ with oxygen deficiency is highly active [17,18], the activity is only observed at the initial stage because the agglomeration of oxygen defects tends to produce amorphous layers at surface that blocks the active site and degrades the catalytic activity in reactions.

Porous single-crystalline β-Ga₂O₃ (PSC Ga₂O₃) monoliths at the macroscale incorporates the synergistic effect of interconnected pores and single-crystalline skeleton in the porous architectures, which yields both long-range ordered and well-defined structures of the lattice in the materials [19,20]. The interconnected pores facilitate the rapid diffusion of species and provide a large surface area for chemical reactions. The long-range ordering of the lattice offers a unique path for creating and stabilizing the V_O in lattice in the interconnected single-crystalline skeletons in the porous architecture. Therefore, PSC Ga₂O₃ monoliths at the macroscale would combine the long-range ordering of the lattice and the disordering of the interconnected pores, thereby improving the properties of the materials themselves. These structures have inherent mechanical and thermal stability, facilitate the species diffusion through 3-dimensional percolation, and are easy to process into complex structures. The monoliths would therefore have the advantages of high catalytic performance and thermal durability in the catalytic reactions [21].

Herein, we fabricate PSC Ga₂O₃ monolith and engineer the oxygen deficiency in lattice to create active sites associated with oxygen vacancies on the well-defined surface in the porous architectures. We further correlate the concentration of V_O with catalytic performance. As a case study, we demonstrate the exceptionally high performance of nonoxidative dehydrogenation of C₂H₆ to C₂H₄ with excellent durability during the continuous operation of 240 h.

PSC Ga₂O₃ monoliths are grown from the mother phase of GaPO₄ single crystal by a lattice reconstruction strategy in a solid-solid transformation process. The growth mechanism is demonstrated in Figs. S1 and S2, which indicates a lattice channel of P/O atoms. The lattice mismatch between β-Ga₂O₃ and GaPO₄ would dominate the preferential growth of facet in the porous architectures. The porosity is ~66% for PSC Ga₂O₃ monoliths, which is favorable for the species diffusion through the 3-dimensional percolation. The x-ray diffraction (XRD) patterns of the PSC Ga₂O₃ monoliths with (100) and (001) facets are grown from GaPO₄ with (−132) and (012) facets, respectively. Figure 1A and Fig. S3 display the microstructures of the PSC Ga₂O₃ monoliths. It is shown that the uniform pore size distribution is at ~100 nm. In Fig. S4, Brunauer–Emmett–Teller test further reveals the pore size of ~110 to 130 nm and specific surface area of ~5 m²/g. Figure 1B and Fig. S5 show scanning transmission electron microscopy (STEM) and selected area electron diffraction (SAED) images of the PSC Ga₂O₃ monoliths, respectively. The porous architecture is pretty uniform with interconnected single-crystalline skeletons. The high-angle annular dark-field image and elemental mappings in Fig. 1C show that the Ga and O are evenly distributed in the skeleton. We have only observed the presence of Ga and O elements in the samples, further proving that the P element is absent, indicating the complete conversion of GaPO₄ to Ga₂O₃ in Fig. S6. The P element has been completely removed during the process by transforming the mother phase of GaPO₄ single crystal into PSC Ga₂O₃ monolith. In addition, the element mapping indicates the absence of element separation in the skeletons. In Fig. 1D, the spherical aberration-corrected high-resolution transmission electron microscopy (Cs-HRTEM) shows the lattice spacings of 0.19 and 0.23 nm, which indicates the good reference to (300) and (202) facets, respectively.

In Fig. 1E and Fig. S7, the high-sensitivity low-energy ion scattering spectra (HS-LEISS) show that the PSC Ga₂O₃ monoliths exhibit the presence of Ga and O on the top layer, forming a series of clear Ga–O coordination structures on the surfaces. According to x-ray photoelectron spectroscopy (XPS) in Fig. 1F and Fig. S8, the oxidation states of Ga and O are +3 and −2, respectively. The ratio of O vacancies calculated from the peak area is only 0.3% to 0.4%, which indicates that the O vacancies can be basically ignored in the pristine PSC Ga₂O₃ monoliths. Time-of-flight neutron powder diffraction (TOF NPD) in Fig. 1G and H indicates that the PSC Ga₂O₃ powder could be indexed assuming a monoclinic with space group C1 2/m. Selected atomic spacings, corresponding to the coordination refined from TOF NPD data, are used to draw the corresponding structure in Fig. 1I. Ga₂O₃ has 2 coordination structures, tetrahedral coordination site (Ga(t)) and octahedral one (Ga(o)). In β-Ga₂O₃ phase, the ratio of Ga(t) to Ga(o) is known as 1:1 [22,23]. The preferential generation of V_O tends to occur in the tetrahedral coordination oxygen ions (O_III sites), which contributes to the unsaturated Ga–O coordination in the monoclinic phase. This suggests that the active sites tend to be produced at the Ga(t) and further indicates that the catalytic activity of oxygen vacancies located in Ga(t) is superior to that in Ga(o).

Figure 2A and B and Figs. S9 and S10 show the formation of different concentration of V_O in lattice, which indicates that the V_O can be isolated even at the concentration of as high as up to 14%. We show detailed information on the density of states of β-Ga₂O₃ with different V_O concentrations. We can observe that the valence band is mainly made up of the O 2p state with a small amount of the Ga 4p state. The local electronic structure of β-Ga₂O₃ also changes with the emergence of V_O [24]. Moreover, the main changes occur in the valence state of oxygen 2p state, and the band gap width of β-Ga₂O₃ with different V_O concentrations becomes narrower compared to pure β-Ga₂O₃. We create the different concentrations of V_O in lattice by directly reducing the PSC Ga₂O₃ monoliths and oxidize the reduced samples by using thermal gravimetric analysis to determine their chemical formula to be Ga₂O_2.952, Ga₂O_2.895, and Ga₂O_2.880, respectively, in Fig. S11. We observe that the XRD of PSC Ga₂O₃ monoliths are slightly shifted to a low angle after reduction, indicating that the introduction of a large number of oxygen vacancies may cause the lattice expansion in Fig. S12A. Nevertheless, no phase transition is observed in the β-Ga₂O₃ even after the high temperature treatment in a reducing gas, firmly verifying the excellent thermal stability of the PSC Ga₂O₃ monoliths. The SAED image of PSC Ga₂O_2.880 monoliths reveals that it is a monoclinic structured β-Ga₂O₃ single crystal with the (100) facet in Fig. S12B, which is consistent with the XRD result. HS-LEISS is used to identify the ratio between O and Ga at the top layer in Fig. S13, which further confirms the changes in surface oxygen concentration from ~25.3% to ~24.8%. Electron spin resonance (ESR) results as shown in Fig. 2C and Fig. S14 further confirm the existence of V_O. The 3 reduced PSC Ga₂O₃ monoliths all show pairs of steep peaks and symmetric distributions at g = 2.003, indicating that electrons are trapped in the V_O [25]. In addition, we further quantify the number of V_O in the reduced PSC Ga₂O₃ monoliths, which gives 2.32 × 10¹³, 2.87 × 10¹³, and 3.45 × 10¹³ mg⁻¹ for the Ga₂O_2.952, Ga₂O_2.895, and Ga₂O_2.880, respectively.

Indirect evidence in Cs-HRTEM is usually utilized to assist the proof of the presence of V_Os especially in domain defects in oxide samples [25,26]. We tentatively conduct the structural analysis based on Cs-HRTEM image to visualize possible defects induced in the samples. The disordered Ga atoms in the form of bright dots in the domain in the size of ~1 to 2 nm have been observed in the bright-field Cs-HRTEM image in Fig. 2D with the inset as marked with an ellipse in the line profile. This may be mainly attributed to the loss of oxygen in lattice that contributes to the formation of domain defects. In contrast, the disordered white clusters that represent O atoms in the domain in the size of ~1 to 2 nm have been observed in the dark-field Cs-HRTEM image in Fig. 2E, which is possibly attributed to the loss of lattice oxygen in the sample, whereas the corresponding domain as marked with an ellipse in the dark-field Cs-HRTEM image in Fig. 2E possesses disordered Ga atoms possibly blocked by V_O defects in the light-field Cs-HRTEM image in the inset. We have added more data to verify the prevalence of domain defects with V_O in Cs-HRTEM images in Fig. S15. This indicates that a large number of oxygen vacancies are randomly distributed in the domain defects in the sample after the high-temperature treatment in a reducing gas. Although the V_O is isolated according to theory calculation, we still observe the slight agglomeration of V_O in lattice. However, the Ga₂O_2.880 still maintains the single-crystalline features, to well stabilize the V_O in the lattice.

To further understand the local structural differences of the PSC Ga₂O₃ monoliths with or without V_O, systematic Ga K-edge extended x-ray absorption fine structure (EXAFS) data are studied by applying Athena-Artemis software in Fig. 2F and Figs. S16 and S17 [27–29]. We observe that there is no phase transformation with or without V_O in the PSC Ga₂O₃ monoliths, and the β-Ga₂O₃ phase is still identified [30,31]. It can be seen that the model well fits with the experimental data. The R-factor is <0.2% in the fitting results with all samples. Actually, the slight variation of coordination number is quite negligible for Ga–O and Ga–Ga of the samples even after reduction because the change of oxygen nonstoichiometry is quite small in Ga₂O_3-δ. Therefore, it is a usual way to fit the EXAFS data by fixing the coordination number of Ga–O in GaO₄ tetrahedra and GaO₆ octahedra at 4 and 6, respectively. All samples are likely to have 1 oscillation in the Ga–O shell. From the fit, the Ga–O bond length is gradually shortened from 1.882 to 1.747 Å for the samples after reduction. Moreover, it is observed that the Ga–Ga bond length is slightly shortened from 3.206 to 3.202 Å for the sample after reduction. We have added the O K-edge EXAFS of PSC Ga₂O₃ monoliths before and after reduction in Fig. S18. The O K-edge spectrum is divided into 2 regions. The first region is between 533 and 544 eV and possesses 2 broad absorption bands. Peaking at 537 and 542.8 eV, respectively, these 2 features arise mainly due to the excitation of O 2p–Ga 4s and O 2p–Ga 4p hybridization orbitals. These 2 polarized peaks tend to shift to lower energies with deeper reduction, and the peaks become shorter due to the shorter bond lengths of the Ga–O bonds. The second region is an EXAFS pseudo-oscillation produced by the multiple scattering effect of the O element with the surrounding metallic elements. This result indicates that V_O can exist stably in reduced PSC Ga₂O₃ monoliths, and these active structures are beneficial for improving dehydrogenation activity and carbon deposition resistance. In Fig. 2G to I and Fig. S19, the XPS results of the O 1s states of the PSC Ga₂O₃ monoliths after reduction indicates the changing trends of the concentration of V_O at surface. Three peaks of oxygen species, fitted from Gaussian simulation, can be attributed to lattice oxygen, oxygen vacancies, and adsorbed oxygen. The chemical state of Ga 2p remains basically unchanged of the 3 reduced PSC Ga₂O₃ monoliths. However, the binding energy of Ga 3d is 19.7 eV in XPS of PSC Ga₂O_2.880 monoliths after reduction. In contrast, the binding energy of Ga 3d is 19.5 eV in the XPS of PSC Ga₂O₃ monoliths before reduction. It can be seen that the binding energy of Ga 3d increases slightly with the increase of the concentration of oxygen defect in lattice. This suggests that the Ga–Ga bond is slightly strengthened, which is consistent with the EXAFS result.

We further utilize ambient-pressure XPS (AP-XPS) in Shanghai Synchrotron Radiation Facility to analyze the chemical changes of Ga and O at surface in the PSC Ga₂O_2.880 monolith under reaction conditions [32,33]. Figure 3A to C shows the spectra of Ga 2p and O 1s of the catalyst in pure ethane from 600 to 660 °C. With increasing temperature, the binding energy of Ga 2p gradually increases by up to ~0.9 eV, indicating that the interaction force of the Ga–Ga bond is strengthened. The asymmetric peak of O 1s in Fig. 3C can be broken into 2 peaks by Gaussian fitting. The stronger peak is attributed to the oxygen in the reduced PSC Ga₂O₃, and the weaker peak is attributed to the C/O species adsorbed on the surface. The binding energy of O 1s is ~1.0 eV higher under a temperature range of 600 to 660 °C, indicating that the higher the temperature is, the easier it is to adsorb C₂H₆ molecules. We then utilize in situ Fourier transform infrared (FTIR) spectroscopy to survey the activation process of C₂H₆ in the PSC Ga₂O_2.880 monoliths in Fig. 3D to F. With increasing temperature, the asymmetric stretching vibration of the C–H bond in ethane is obviously shifted to the blue region, indicating that the C–H bond is effectively activated to a higher energy state. Additionally, the stretching vibrations of the C–H and C=C bonds of C₂H₄ become increasingly obvious, suggesting that the initial temperature of C₂H₆ dehydrogenation to C₂H₄ in the β-Ga₂O₃ phase is as low as ~200 °C. The CH₂, C=CH₂, and OH are the intermediates in the process of the nonoxidative dehydrogenation from C₂H₆ to C₂H₄ [34].

Information about the symmetry of the crystal structures is obtained by the Raman analysis, which indicates the monoclinic structure of β-Ga₂O₃. It consists of a tetrahedron structure (Ga_IO₄) and an octahedron structure (Ga_IIO₆) and includes 2 categories of Ga atoms (Ga_I and Ga_II) and 3 categories of O atoms (O_I, O_II and O_III) in Fig. S20. We utilize in situ Raman spectra to survey the phase transition of C₂H₆ in the PSC Ga₂O_2.880 monoliths in Fig. 3G. The Raman peak confirms that no phase transition occurred [35]. Figure S21 shows the FTIR spectroscopy of adsorbed pyridine for distinguishing between a Lewis acid and Bronsted acid. The absence of Bronsted acid in the β-Ga₂O₃ phase is indicated by the absence of a characteristic absorption band at ~1,540 cm⁻¹. Figure 3H shows the configuration of chemisorption of C₂H₆ at surface, which indicates that the electrons are transferred from H to Ga atoms in the configurations. Figure 3I shows the process (C₂H₆ (g) → C₂H₆* → C₂H₅* + H* → C₂H₄* + 2H*) and the density functional theory (DFT)–calculated potential energy for the direct dehydrogenation of C₂H₆. The adsorption energy (Eads) of the C₂H₆ molecule is −0.23 eV, and the energy barrier is 1.56 eV for each elementary process. Reasonable activation energies and potential energy barriers indicate that the adsorbed C₂H₆ molecules can be efficiently activated, and the C–H bond cleavage tends to convert C₂H₆ to C₂H₄.

Figure 4A and Fig. S22 display the product analysis for the nonoxidative dehydrogenation from C₂H₆ to C₂H₄. We discover that the ethylene is a dominant product and that the methane exists as a by-product. This result suggests a carbon balance of ~100% in Fig. 4B, further indicating that there is no carbon deposition during the reaction. The well-defined surfaces coupled with tailored V_O effectively improve the activation of C–H bonds in C₂H₆ and coking resistance. Figure 4C and D and Fig. S23 display the C₂H₆ conversion and C₂H₄ selectivity during the nonoxidative dehydrogenation process. C₂H₆ conversion on PSC Ga₂O_2.880 monoliths gradually increase versus temperature and finally reaches ~37% at 660 °C. In the control experiment, it is ~5 times higher than that of the thermal cracking of C₂H₆ without catalyst. The C₂H₄ selectivity of the PSC Ga₂O_2.880 monoliths can be up to ~100%. That of the control experiment is much higher than the ethylene selectivity of the thermal cracking of C₂H₆ without a catalyst. Here, the well-defined surfaces coupled with tailored V_O are considered to be active sites for the dehydrogenation of C₂H₆. The property of the active site governs the difference in the catalytic conversion and selectivity for dehydrogenating C₂H₆ to C₂H₄. In addition, the C₂H₆ dehydrogenation of different samples is performed at 620 °C, and stable C₂H₆ conversion and C₂H₄ selectivity are shown in Fig. 4E and F without degradation after 240 h of operation. We have added the characterization of the catalysts after the stability test for the dehydrogenation of C₂H₆ at 620 °C in Fig. S24. The XRD patterns confirm the structural stability of the 3 reduced PSC Ga₂O₃ monoliths after the stability test. The chemical states of Ga and O remain basically unchanged after the stability test. The Raman spectra confirm no visible coke formation after the stability test. In addition, the microstructure of the PSC Ga₂O_2.880 monoliths generally remain unchanged after stability test in Fig. S25. The domain defects associated with oxygen vacancies are not increased in lattice, which also reveals the excellent structural stability of PSC Ga₂O_2.880 monolith.

In order to further validate the stability, we conduct in situ spectroscopy to understand the surface states of the PSC Ga₂O_2.880 monoliths during the reaction. In Fig. 4G, we perform in situ Raman scattering studies of the PSC Ga₂O_2.880 monoliths from 0 to 100 h in pure C₂H₆ and find that the crystal structure remains stable without any visible change. In addition, we also find no obvious coke formation during the 100-h test. We perform in situ HS-LEISS spectra of the PSC Ga₂O_2.880 monoliths from 0 to 50 h in pure C₂H₆ in Fig. 4H. The compositions of Ga and O at the top layers remain essentially constant, and there is no carbon peak on the surface, indicating that the PSC Ga₂O_2.880 monoliths can exist stably under pure C₂H₆ at high temperature. These results indicate that the well-defined surfaces of the PSC Ga₂O₃ monoliths can accommodate oxygen vacancies, thus stabilizing the crystal structure and improving the carbon deposition resistance. Figure 4I presents the in situ XPS spectra of the PSC Ga₂O_2.880 monoliths after 0 to 50 h of reaction in pure C₂H₆. The binding energy peaks of Ga 3d at 22.96, 19.65, and 18.59 eV are visible, which originates from the O 2s, Ga–O, and Ga–Ga bonds of the reduced PSC Ga₂O₃, respectively. Moreover, the Ga 3d peak intensity is not visibly changed with increasing reaction time.

In summary, we quantitatively introduce V_Os and stabilize them by the long-range ordering of crystal lattice in a centimeter-scale PSC Ga₂O₃ monolith to boost the catalytic activity and durability. The oxygen vacancies are well stabilized in lattice even with slight agglomeration into domain defects ~1 to 2 nm at a high oxygen deficiency. The formation of low-coordinated Ga active sites therefore delivers an exceptionally high C₂H₄ selectivity of ~100% at an C₂H₆ conversion of ~37% even only at 660 °C for the nonoxidative dehydrogenation of C₂H₆ to C₂H₄. We have further demonstrated the outstanding stability in a continuous operation of up to 240 h. This work would not only facilitate the rational design of catalytic materials but also find potential applications of the porous single crystals in other fields.